Why Better Dashboards Don't Always Lead to Better Business Decisions
A better-looking dashboard is not the same as a better decision. Why reporting improvements often stall, and what to design for instead.
Most organisations that invest in business intelligence end up with better dashboards. Fewer end up making noticeably better decisions. The gap between the two is not a technology problem, and it is rarely closed by another redesign of the visuals.
In my work building reporting models, the most common pattern I see is a dashboard that is accurate, well designed and opened regularly — and that changes very little about what anyone actually does. It is worth asking why.
A dashboard answers questions. Someone has to ask them.
A dashboard is a set of pre-answered questions. If those questions were chosen because the data was available, rather than because a decision depended on them, the result is a tidy display of facts nobody needed.
The useful starting point is not “what data do we have?” but “what decision is made, by whom, how often, and what would change it?” A regional manager deciding where to spend next week’s time needs to know which stores are drifting from target and whether the drift is new. They do not need fourteen charts of last year’s totals.
When I design a report, I try to write down the decision in one sentence before choosing a single visual. If I cannot write that sentence, the report is not ready to be built.
Trust comes before insight
The second reason dashboards fail is that people do not fully trust them. Trust is lost quickly and recovered slowly. One visible mismatch between the dashboard and a figure someone checked by hand is enough for a manager to start keeping their own spreadsheet “just in case” — and from then on, the dashboard is decoration.
Trust is built in the parts of a report users never see: reconciliation against source systems, clear handling of missing data, and consistent definitions. A sales figure that silently excludes late submissions is not wrong exactly, but it will eventually disagree with someone’s own records, and the damage will be done.
In practice this means spending a large share of effort on the unglamorous work: agreeing definitions, reconciling totals, and showing data completeness on the report itself. A small indicator that says “three stores have not submitted yet” does more for trust than any amount of visual polish.
Totals hide the things that matter
Executive reporting tends towards totals and growth rates because they are easy to read. But many important business facts live in distributions, not totals.
Revenue can grow while becoming dangerously concentrated in a handful of customers. Average processing time can look stable while a growing share of cases wait far longer than they should. Neither shows up in a headline number.
Good decision-support reporting deliberately includes views of distribution and exception: who contributes most, where the outliers are, and what has changed since last period. These views are harder to design than a total, but they are usually where the decision actually sits.
Reports describe; people still have to act
Even a relevant, trusted, well-structured report only describes a situation. The decision still needs an owner and a moment when it is made. Reporting that is not connected to a recurring decision forum — a weekly operations meeting, a monthly performance review — tends to drift into being “available” rather than used.
The question I find most useful at handover is simple: when will someone next look at this, and what will they do differently because of it? If the answer is vague, the work is not finished, however good the dashboard looks.
What to design for instead
If better dashboards are not enough, what should reporting work aim for? In my experience, four things:
- A named decision. Each report or page supports a specific, recurring decision, written down before the build.
- Evidence of reliability. Reconciliation to source, explicit data-completeness indicators and stable definitions.
- Exceptions over totals. Views that show what is unusual, concentrated or changing, not only what is large.
- A place in the operating rhythm. A meeting, owner or routine where the report is used, and where its usefulness is reviewed.
None of these is primarily about visual design. Visual design still matters — a confusing chart is a real cost — but it is the last step, not the first.
Closing thought
The value of business intelligence is measured in decisions, not in dashboards. A plain report that a manager trusts and uses every Monday is worth more than a beautiful one they glance at and set aside. Designing for the decision, and for trust, is slower at the start. It is also the only approach I have seen consistently pay off.